基于YOLO v5的施工现场清洁度智能识别与评定

Intelligent Identification and Assessment of Cleanliness of Construction Site Based on YOLO v5

针对施工现场灰尘识别缺乏积灰程度的识别,本文提出了基于YOLO v5算法的灰尘检测与清洁度评定方法,研究选择高效的YOLO v5算法对几种工地灰尘和垃圾进行检测并分配不同的权重系数,统计检测目标数量加权出清洁度指数,进而对检测区域清洁度作出等级评定。系统根据清洁度等级来指挥平面清洁机器人进行清洁任务,使其高效完成清洁作业。实验结果表明:模型检测准确率为90.6%,灰尘和工地垃圾在测试集上的mAP@0.5达到90.9%。本文算法可准确灰尘识别并对检测的工地平面进行清洁度评定。

In this paper,in view of the lack of identification of ash accumulation degree in construction site dust identification,a dust detection and cleanliness evaluation method based on YOLO v5 algorithm was proposed.The efficient YOLO v5 algorithm is selected to detect dust and garbage on several construction sites and assign different weight coefficients.The cleanliness index is weighted by counting the number of detection objects,and then the cleanliness level of the detection area is evaluated.The system directs the plane cleaning robot to perform cleaning tasks according to the level of cleanliness,so that it can efficiently complete the cleaning operation.The model detection precision is 91%,and the mAP@0.5 of all kinds of construction dust and site waste on the test set reaches 90.9%.It is confirmed that the algorithm in this paper can accurately identify dust and evaluate the cleanliness of the detected construction site.